The present invention relates to a mobile communication system operating on what is known as the code division multiple access (CDMA) system.
The CDMA system involves multiplexing a plurality of communication channels using spread spectrum codes, each channel being assigned a different spread spectrum code. A given signal to be transmitted is multiplied (i.e., spread) by the spread code assigned to the own channel, and is multiplexed with other similarly spread signals on different channels before being transmitted. At a receiver, the multiplexed signals are multiplied (i.e., despread) by the same spread code so that only the target signal will be extracted correlated on the own channel. The signals on the other channels are perceived merely as noise because these signals with their different spread codes remain uncorrelated. The level of the noise may be sufficiently lowered so as not to disturb the signal reception. The CDMA system is attracting attention as a system fit for drastically improving the efficiency of frequency utilization and has been commercialized in some areas.
Where CDMA communication is implemented using spread codes, some kind of signal modulation (e.g., quadrature phase shift keying or QPSK) precedes the spreading of the signal for transmission. At a receiving point, the despreading of the signal is followed by demodulation. Despreading and demodulation both represent the detection process whereby the transmitted signal is reconstructed. Commonly used detection methods include a coherent detection method based on the PLL (phase locked loop) circuit and a differential detection method. There also exists a recently proposed coherent detection method that utilizes pilot signals.
Where the CDMA system is applied to a mobile communication system adopting the conventional coherent detection method, the bit error rate of data in a mobile station deteriorates if a fading occurs while the station is moving. In a CDMA mobile communication system utilizing the differential detection method, the bit error rate of data in a mobile station can worsen due to the noise on the air transmission channel even if the station is stationary. The pilot signal-based coherent detection method has been proposed for a system to minimize the deterioration of the bit error rate whether the mobile station is in motion or at rest. The method was discussed at the Autumn 1994 Symposium of the Institute of Electronics, Information and Communication Engineers of Japan as disclosed in the IEICE collection of papers B-5 on radio communication systems A and B, p. 306, "Coherent detection for CDMA Mobile Communication Systems" by Yasuo Ohgoshi et al.
Described below is a conventional mobile communication system that uses pilot signals with reference to the above-cited paper supplemented by some details. The description will first center on the down link of the system (i.e., a link from the base station to a mobile station). FIG. 13 shows a modulation circuit 51 of a base station 1 that transmits data and a first half 52 of the detection circuit of a mobile station 2. The base station 1 actually transmits signals to a plurality of mobile stations 2, and FIG. 13 shows one station as the representative example.
In the modulation circuit 51 (left-hand half of FIG. 13), data first undergoes QPSK modulation, not shown, to divide into an in-phase signal I and a quadrature signal Q. The signals I and Q are spread (i.e., multiplied) respectively by spread code signals PN.sub.-ID and PN.sub.-QD. The two spread code signals are supplied from a spread code generator 91. The rates of the spread code signals PN.sub.-ID and PN.sub.QD (called the chip rates) are used to multiply by k (k: spreading ratio) the pre-spread rates (called the symbol rates) of the signals I and Q so that the latter will attain the chip rates. The signals thus spread pass through a radio frequency quadrature modulator 54 to become mutually perpendicular signals that are transmitted on a radio frequency band from an antenna. A temperature compensated crystal oscillator 61 is provided to furnish the modulator 54 with a carrier C.sub.B.
The pilot signals will now be described. The transmission circuit is substantially the same as the left-hand half of FIG. 13 and is omitted. An in-phase signal I.sub.P and a quadrature signal Q.sub.p of the pilot signals are spread respectively by spread code signals PN.sub.IP and PN.sub.-QP. Both spread code signals have the same chip rate as in the case of data. The pilot signals thus spread are subject to radio frequency quadrature modulation by the same carrier C.sub.B as with data, turning into mutually perpendicular signals transmitted on the same radio frequency band as with data. The pilot signals serve as reference signals for demodulation and are common to all channels utilized.
In the first half 52 (right-hand half of FIG. 13) of the detection circuit of the mobile station 2, the received signals from the antenna (data and the pilot signals) pass through a radio frequency quadrature demodulator 57 to reach a low-pass filter 56. The low-pass filter 56 removes the radio frequency components from the signals to yield signals S.sub.I and S.sub.Q. A crystal oscillator 60 supplies the demodulator 57 with a carrier C.sub.M. The signals S.sub.I and S.sub.Q are composed of the spread signals I and Q (those destined to the own channel as well as to other channels) and of the spread pilot signals I.sub.P and Q.sub.P. As such, the signals S.sub.I and S.sub.Q include a phase error caused by fading and a frequency error attributable to the precision of the oscillator 60.
The errors included in the signals S.sub.I and S.sub.Q produce a phase difference therein. When the mutually perpendicular pilot signals are plotted in orthogonal coordinates, the received pilot signals are rotated exactly by the phase shift, as shown in FIG. 14. If the phase shift is represented by .phi. and the orthogonal coordinates after quadrature demodulation are designated by X.sub.1 and Y.sub.1, then the coordinate axes X and Y of the received signals are rotated by .phi. displacing the pilot signals. Consequently, the undisplaced signals i and q that should have resulted with no phase shift become i.sub.1 and q.sub.1 respectively. Such changes are caused by the mixing of one of the two mutually perpendicular signals into the other signal. The phenomenon is expressed by the following formulas: EQU i.sub.1 =i cos .phi.-q sin .phi. EQU q.sub.1 =q cos .phi.+i sin .phi.
The pilot signals are signals that stay constant following the despreading. Generally, i=1 and q=1. The signal changes into i.sub.1 and q.sub.1 permit acquisition of a signal CS with the value cos .phi. and a signal SN with the value sin .phi.. With the two signals known, it is possible to correct the phase rotation of the data. Since the data includes the same phase shift, the despread data signals are inversely rotated by .phi. using the signals CS and SN whereby the initial signals I and Q are correctly reconstructed. Thus the signals CS and SN serve as phase correction signals.
The signals S.sub.I and S.sub.Q output by the first half 52 of the detection circuit are subject to despreading and phase correction by the second half of the detection circuit shown in FIG. 15. A pilot signal despreading unit 21 in the upper left portion of FIG. 15 despreads the signals S.sub.I and S.sub.Q by use of the spread code signals PN.sub.-IP and PN.sub.-QP from a spread code generator 25, whereby the pilot signals are extracted. The extracted pilot signals are then added and subtracted mutually, becoming a signal CS.sub.C with a chip rate of cos .phi. and a signal SN.sub.C with a chip rate of sin .phi.. The two signals are converted to the symbol rates by an accumulator 41 and thereby turn into phase correction signals CS.sub.S and SN.sub.S of the preliminary stage. The phase correction signals are averaged by an averaging circuit 43 for noise reduction. The averaging provides the phase correction signals CS and SN of the final stage.
FIG. 16 shows a typical circuit constitution of the averaging circuit 43. Reference numerals 430 through 433 are delay gates (Ds) for delaying a signal by a one-symbol period each. In this example, three consecutive symbol values are averaged when added up by adders 235 and 236. It is through this noise reduction arrangement that the phase correction signals CS and SN are obtained. The delay time (average delay time) T required for the averaging by the averaging circuit 43 is given as EQU T=Ds.times.(N-1)/2
where N denotes the number of symbols used for the averaging operation.
The data signals S.sub.I and S.sub.Q are both despreads by an inverse data spreading unit 42 (bottom left in FIG. 15) using the spread code signal PN.sub.-ID for the signal I and the spread code signal PN.sub.-QD for the signal Q. The operation causes four signals to be extracted. The four chip rate signals are converted by an accumulator 44 into symbol rates to become signals D.sub.1 through D.sub.4. After this, the signals D.sub.1 through D.sub.4 are each delayed by a data delaying unit 48 (FIG. 17) by the average delay time T of the averaging circuit 43. The operation yields signals D.sub.10 through D.sub.40. Where the data delaying unit 48 is constituted by a number of delay gates (Ds) in stages of cascade connection each gate providing one-symbol period delay, the gate count M per stage is given as EQU M=(N-1)/2
In the above example, N=3 and thus M=1, so that the delay gates 480 through 483 of the data delaying unit 48 are each composed of a one-symbol delay gate (Ds).
The signals D.sub.10 through D.sub.40 are fed to a phase correction circuit 49 in which the signals are corrected in phase rotation by use of the correction signals CS and SN. A typical constitution of the phase correction circuit 49 is shown in FIG. 18. The phase correction circuit 49 performs phase correction as follows: the signals D.sub.10 and D.sub.40 are multiplied by the correction signal CS, and the signals D.sub.20 and D.sub.30 by the correction signal SN. The multiplied results are added and subtracted mutually so as to rotate the orthogonal axes of the received data by -.phi. in phase (i.e., the phase shift .phi. is reduced to zero in FIG. 14). The phase correction provides reconstructed signals I.sub.R and Q.sub.R of the original signals I and Q. The signals I.sub.R and Q.sub.R then undergo QPSK demodulation, not shown, to become the original data.
One disadvantage of the conventional detection circuit above is that the restored signals I.sub.R and Q.sub.R are unavoidably affected by the frequency precision of the crystal oscillator 60 (right-hand side in FIG. 13). A transmitter 60 used in the mobile station necessarily includes a certain practical frequency error because the mobile station is for use by general users. That is, on the one hand, if the frequency error involved in the data is large enough to cause apparent phase irregularities over the average delay time T during data demodulation, no precise correction signals can be acquired and the bit error rate of the detected data worsens. On the other hand, if the average delay time T is shortened to avert the deterioration of the bit error rate, the adverse effects of the frequency error are diminished but the line noise becomes more pronounced.
On the up link (i.e., a link from the mobile station to the base station), the carrier C.sub.M from the crystal oscillator 60 often doubles as a carrier for use in radio frequency quadrature modulation by the modulation circuit of the mobile station. In that case, the signals transmitted by the mobile station and received by the base station include both the phase error caused by fading and the frequency error originating from the crystal oscillator. The frequency error results in the inevitable deterioration of the bit error rate in the detection process of the base station.
The deficiencies above are conventionally circumvented, particularly where data of lower bit rates than normal are transmitted, by the method of burst data transmission with no change in the spreading ratio, as stipulated by the U.S. digital radio communication standard IS (Interim Standard)-95. Under the system, transmitting data at 1/r of the standard bit rate compresses the data to l/r in temporal terms. The time-compressed data is transmitted in bursts at fixed intervals.
How the burst signals are sent intermittently is illustrated in FIG. 19. In FIG. 19, the axis of abscissa represents time and the axis of ordinate denotes transmission power. Reference numeral 140 is a radio signal waveform of standard bit rate data. Reference numerals 141, 142 and 143 stand respectively for radio signal waveforms of data at 1/2, 1/4 and 1/8 of the standard bit rate. The number of burst signals varies with the bit rate. All burst signals have the same standard bit rate when temporally compressed as described. It follows that every burst signal has the same symbol rate and thus the spreading ratio remains unchanged.
The arrangements above are necessitated by the following reasons: if compression is not carried out, the one-symbol period gets longer the lower the data rate. Meanwhile, the number of symbols N for use by the averaging circuit 43 (FIG. 15) in the demodulation circuit remains substantially the same regardless of the bit rate in view of noise reduction. Thus the average delay time T becomes longer the lower the data rate. A prolonged delay time T prompts the frequency error to deteriorate the bit error rate as discussed above. The lower the bit rate, the more deteriorated the bit error rate. To avoid this deficiency requires keeping the symbol rate constant. The requirement necessitates the use of complicated circuits in the mobile station, which runs counter to the inherent need for the mobile station to simplify its circuitry.